In a large-scaled data management system, one server may have data of documents and the like that are frequently accessed from many users. In this case, to distribute the load, generation of a plurality of apparatuses called mirror servers for holding identical data contents has been proposed. When such mirror servers are used, a user who requires a document or the like on the original server can extract the document or the like from an appropriate mirror server, and the load can be distributed.
However, even when mirror servers are installed, most users cannot determine which mirror server can be efficiently accessed. In addition, when mirror servers are generated, the maintenance cost becomes high.